来源：https://dev.to/michaeltruong/the-agent-plan-had-every-step-except-where-to-stop-357h

I've been running multi-slice agent plans — Renovate migrations, content-pipeline skills, dependency upgrades. I split multi-PR work into slices, each backed by a markdown file with file paths, verification commands, and merge-safe acceptance criteria.

I assumed the checklist was enough. The plan described what to build. I treated how far the agent could go as implicit. Then an agent merged a pull request I expected to review first.

The trigger was mundane. During the first slice of a Renovate migration, an agent regrouped dependency buckets in renovate.json — config-only, no version bumps, no runtime behavior. It ran lint and typecheck, opened the pull request, and merged it. The change itself was reasonable.

What surprised me was the absence of a documented stop line. The migration plan described the edit, the verification commands, and the acceptance criteria. It did not say whether the executing agent should stop at "open PR" or continue to "merge after green checks." The plan was an implementation spec. The agent treated it as permission to finish the job.

Traditional engineering plans answer: what work should happen, in what order, with what verification? Agent plans increasingly need a second answer: how much autonomy does the next actor get?

My first reaction was to tighten the repository boundary — branch protection became the safety layer. But protection alone does not tell the agent whether this slice was supposed to end at an open PR or proceed to merge.

The portable fix: every slice names exactly how far the executor may go. Two levels: Default is Open PR only. Elevated is Merge granted, requiring explicit rationale. Each slice also states Rationale and copies the Agent instruction verbatim into the prompt.

The lesson is narrower: once agents act, plans delegate autonomy whether you write that down or not. Human delegation has always been fuzzy — "take a pass at this" means different things to different people. Agent delegation punishes ambiguity faster because the agent will complete every step it can justify from the text in front of it.

（此帖无评论）